Research On Tool Wear State Monitoring Technology Of High Speed Milling Based On Deep Learning

As the increasing development of automation and intelligent during the machining process,the research on mechanical fault intelligent monitoring technology has become crucial.Cutting tool in machining is one of the most important processing elements,since tool wear not only directly affect the dimensional accuracy and surface quality of the workpiece,but also indirectly affect the machining efficiency and production cost,etc.Now how to use more effective methods to monitor tool wear has become a large number of scholars’ research priorities.Research on state monitoring technology of tool wear has huge potential and application value.This dissertation is based on the state of tool wear in high speed milling process as the research object,and then study the different monitoring technology with the cutting tool wear in different condition of milling force signal.The whole dissertation is organized as follows:(1)the establishment of a state monitoring test system with high speed milling tool wear,through the test of pressure sensors which collect a large amount of force signal data,(2)choosing 13 typical features as the input vector to the neural network in the following step,based on the feature extraction and feature selection of the milling force signals in time domain,frequency domain and time-frequency domain,(3)setting up a model of tool wear condition monitoring based on BP neural network,(4)proposed a deep study of high speed milling tool wear state monitoring methods,via introducing the deep learning theory into the field of tool wear monitoring.The main contribution to this dissertation is to propose a deep learning method for realizing the intelligent monitoring of tool wear.Firstly,the wavelet packet transform is employed to extract the milling force signals in different frequency band energy distribution as the initial feature vector.Secondly,the feature in sparse coding network is learned through an unsupervised learning method.Next,single neural networks form deep network.Finally,in the light of supervised learning on the depth of the whole network for fine-tuning training,the intelligent monitoring model of the milling tool wear is established.The experimental results show that compared with the traditional model,the proposed new method based on the theory of deep learning on tool wear monitoring can adaptively extract new information of tool wear,and this method has higher accuracy of monitoring.